2012/9/26 Lester Caine <[email protected]>: > Christian Quest wrote: >> >> So cross check with Bing >> must be done afterwards, exactly like when using vector data. >> That's why I consider manual tracing as a waste of time, and not high >> quality compared to using extracted building from vector data. > > > Christian > I've now seen your source data, and that is showing buildings as 'blocks'. > In the PDF file are those blocks drawn individually when there are are more > than one building side by side? Or are the vectors simply the outline of the > whole block? >
There is no single answer as there is many different cases. Most of the time, a building is defined by a single polygon. Sometimes, in may be split into multiple polygons if the land/building ownership has evolved during the past or if the building is composed of different parts. Auto-merging is almost impossible and manual merging is not obvious. Sometimes, one building polygon is covering different building parts. Auto-split is also almost impossible and manual split also not obvious. > I'm sure that over time improved quality data will evolve, but the current > vector data that many of us have access to is not ideal. I doubt we will ever have access to a better source nation wide. Some cities that stepped into opendata do provide building vector data but they are a minority and most of the time it is the same data because cities are updating the cadastre data locally then send it to the nationwide administration. Cities and town had/have the responsibility to create the vector version. Some parts of France are fully vectorized, some are still very late (25% of our 36000+ communes still). > Personally the > problems I am finding is where we have semi detached houses drawn as a > single block, and splitting that into two blocks is a pain on potlatch ... Well... working with complex objects may be a pain in Potlatch but this is not typical of building coming from the cadastre. Some building with wholes in them are even using multipolygon relations. > I've not tried on JOSM yet. But a tool on my 'wish list' is one I can use to > select vector lines from a 'staging layer' and combining them to a closed > way to which I can then add extra tags. This I think is the best way to use > this 'poor quality' vector data and convert it to better quality data? I can > also see a 'split' tool, where the imported vectors are for a 'semi' or > 'terrace' and you want to split each out to separate buildings. > I agree that a tool in JOSM (or whatever is your editor of choice) to split a polygon into 2 smaller polygons could be really helpful, not only for buildings. > http://www.openstreetmap.org/?lat=52.048913&lon=-1.85734&zoom=18 will show > my own experience with all of this. What is not visible is the Opendata > streetview layer, and just how bad the vector data is with respect to the > current status. 'New' buildings are actually not too bad, but none of the > extensions on the houses on Smallbrook Road are present on the OS layer, > which seems to be stuck with 40+ year old data. Some how I expect the same > sort of discrepancies in most data, and so I would not use the OS data in > the same manor you are using the French data, although with my historic hat > on it WOULD be nice to retain the history of the additions of these > extensions over time. Importing buildings from the Opendata streetview layer > would fill up the UK map, but we do not know what date the buildings relate > to and then updating and we still need to split and add address data ... > The cadastre is not perfect, but it is not 40+ years old data, it is closer to 1 or 2 years old in most cases. The updates are available more or less once a year on a town by town basis. Cadastre is used to collect taxes and our administration is quite efficient in that area ;) In the source tag, "dgi" means Direction Générale des Impôts (Impôts = taxes). What's missing from time to time in the cadastre data are state/public buildings as they don't pay taxes ! One additional thing we have been able to extract from vector cadastre data is a rough length of highways/path/tracks in each town. This allows to have a rough idea of the existing/missing ratio as displayed by this layer: http://layers.openstreetmap.fr/?zoom=7&lat=47.45473&lon=2.19472&layers=B00FFFFFFFFFFFFFFFFFT -- Christian Quest - OpenStreetMap France - http://openstreetmap.fr/u/cquest _______________________________________________ talk mailing list [email protected] http://lists.openstreetmap.org/listinfo/talk
